Physics-Based 3D Simulation for Synthetic Data Generation and Failure Analysis in Packaging Stability Assessment
The design and analysis of pallet setups are essential for ensuring safety of packages transportation. With rising demands in the logistics sector, the development of automated systems utilizing advanced technologies has become increasingly crucial. Moreover, the widespread use of plastic wrapping has motivated researchers to investigate eco-friendly alternatives that still adhere to safety standards. We present a fully controllable and accurate physical simulation system capable of replicating the behavior of moving pallets. It features a 3D graphics-based virtual environment that supports a wide range of configurations, including variable package layouts, different wrapping materials, and diverse dynamic conditions. This innovative approach reduces the need for physical testing, cutting costs and environmental impact while improving measurement accuracy for analyzing pallet dynamics. Additionally, we train a deep neural network to evaluate the rendered videos generated by our simulator, as a crash-test predictor for pallet configurations, further enhancing the system's utility in safety analysis.
Code (0)
등록된 구현이 없습니다.
Tasks
Synthetic Data GenerationSimilar Papers 제목 키워드 기반
Controlled physics-informed data generation for deep learning-based remaining useful life prediction under unseen operation conditions
Limited availability of representative time-to-failure (TTF) trajectories either limits the performance of deep learning (DL)-based approaches on remaining useful life (RUL) prediction in practice or even precludes their…
Generative Adversarial NetworkNewtPhys: Do Foundation Models Understand Newtonian Physics?
Previous work has evaluated physics reasoning in foundation models using synthetic or semi-synthetic scenes and visual question-answering tasks. However, these benchmarks emphasize high-level events and lack the visual f…
Synthetic Video Enhances Physical Fidelity in Video Synthesis
We investigate how to enhance the physical fidelity of video generation models by leveraging synthetic videos derived from computer graphics pipelines. These rendered videos respect real-world physics, such as maintainin…
Video GenerationSIM1: Physics-Aligned Simulator as Zero-Shot Data Scaler in Deformable Worlds
Robotic manipulation with deformable objects represents a data-intensive regime in embodied learning, where shape, contact, and topology co-evolve in ways that far exceed the variability of rigids. Although simulation pr…
Your Simulation Runs but Solves the Wrong Physics: PDE-Grounded Intent Verification for LLM-Generated Multiphysics Simulation Code
Execution-based evaluation of LLM-generated code implicitly treats successful execution as a proxy for correctness. In scientific simulation, this proxy is insufficient: a generated input file can run, mesh, and converge…